Ensemble-based estimation of German methane emissions using the ICON-ART model

January 07, 2026

In the ITMS project, a new system is being developed at the DWD to assess greenhouse gas emissions based on concetration measurements and weather modeling. Now we have published two new papers in the journal Atmospheric Chemistry and Physics, describing the ITMS-Demonstratos and estimating the Germany's methane emissions in 2021. Here, we explain the foundations of this system.

Introduction

Methane (CH4) is among the most important greenhouse gases. After carbon dioxide (CO2) it is the gas with the strongest impact on the climate. The reduction of methane emissions is therefore crucial in climate change mitigation efforts.

In Germany, most of the emissions are from the agricultural sector, followed by public power and waste management. These emissions are collected yearly into national greenhouse gas emission inventories and submitted to the UN. The Umweltbundesamt (UBA) is responsible for this task in Germany. For policy makers, the accuracy of these numbers is essential, because they document the efficacy of the emission reduction efforts. Germany has committed itself to reduce its methane emissions by at least 30% until 2030 compared to 2020.

The ITMS project aims to infer the German greenhouse gas emission based on atmospheric observations. This top-down approach allows to estimate the emissions based on observations of the greenhouse gas concentrations in the atmosphere. In this way, it can assist the bottom-up approach of the national greenhouse gas emission inventories, and help to verify them.

The recent scientific publication consists of two parts and is published in the journal EGUSphere. The first part describes the method, while the second part presents the results. Together, they represent the first application of the top-down approach for Germany in 2021 by the ITMS-Demonstrator.

What was done?

Methane transport

To simulate the methane emissions and concentrations, the weather prediction model of the German Meteorological Service is employed: ICON. The model calculates how air parcels move, turbulence arises, and how a gas spreads in the atmosphere. In this way, the methane emissions can be traced.

The weather is a chaotic system, in which small changes in the initial conditions can lead to large changes in the prediction. To mitigate this, not one simulation is utilized, but a set of twelve slightly different versions of the weather — a so-called ensemble. This allows an estimation of the uncertainties in the weather transport model.

Emissions

Before starting the simulations, an initial estimate of the emissions is needed — the prior. This is the initial best guess that the method starts with. For Germany, these numbers come from the greenhouse gas inventory of UBA. For the rest of Europe, the anthropogenic emissions are taken from the Copernicus Atmosphere Monitoring Service.

There are also natural sources of methane, such as wetlands and permafrost thawing. The data for Germany is obtained from the Thünen Institute, but these emissions are comparatively small. The natural fluxes become more relevant in Northern Europe.

*Figure 2: ICOS tower at Gartow*

*Figure 2: ICOS tower at Gartow*

ICOS-Observations

To compare the models with reality, observations of the methane concentrations are needed. The Integrated Carbon Observation System (ICOS) is a European network of observation stations, which monitor the concentration of greenhouse gases continuously.

There are more than a hundred observational towers, some of which can reach hundreds of meters in height.

The observations follow high quality standards and are publicly available through the ICOS Carbon Portal. They are essential for the inversion, because without accurate measurements, the emissions cannot be verified.

Inversion

*Figure 3: The 46 emission categories colored. The dots mark the sites of the observation stations*

*Figure 3: The 46 emission categories colored. The dots mark the sites of the observation stations*

The inversion is at the heart of the scientific publication. The crucial question is calculated backwards: If we know where the wind blows and how much methane is at the observational site, where was it emitted?

*Figure 4: The simulated contributions of the emission categories at one observation site, compared to the measurements*

*Figure 4: The simulated contributions of the emission categories at one observation site, compared to the measurements*

The approach collects the emissions into regions and sectors into categories. Figure 3 shows a map of them. For each category, the model simulates the methane concentrations separately. At the observational sites, the simulated observation can then be broken into its constituent categories (while the real observation only measures the total methane). If the simulated observations of the initial guess do not match the real observations, we can scale the categories by a multiplicative factor to get a better fit. For a single observation, there are many ways to scale the categories; mathematically this problem is underdefined. But the study uses 100 000 hourly observations throughout the year 2021. In this way, the best fit for the categories can be calculated on average. This method gives, for example, a scaling factor for German emissions of 1.32 ± 0.17, which means an increase of 32% ± 17% of the emissions — more on this in the next section.

Results

*Figure 5: Top-down estimated methane emissions in 2021*

*Figure 5: Top-down estimated methane emissions in 2021*

The results of the inversion are shown in Figure 5.

For Germany, the method gives a 32% ± 17% increase in the total methane emissions compared to the national inventory (Version 2024). In the Netherlands the method also indicates higher than expected emissions, while in Great Britain it is in accordance with the inventory.

Challenges

There are limitations to the method. Outside Central Europe, the coverage of observations is not sufficient to use the method effectively.

The collection of emissions into categories ensures that whole regions are corrected together. If only a part of the region should be corrected, the method cannot resolve it. For several sectors in the same region — as in Germany — the capacity for differentiation is limited. It is easier to estimate the total emissions of a region. For example, in Germany, the method attributes most of the emission increase to agriculture, but this is more uncertain than the increase in the total emissions.

In addition, the method shows a changing seasonal cycle: The emissions peak in fall and winter, and reach a minimum in May. This is different in comparison to other transport models, which could indicate inaccuracies in the transport models. These inaccuracies generally present the largest challenges of the top-down inverse methods.

Outlook

The code of the inversion will be publicly available as open source and can be found here (see also: DUBFI)..

The list of possible further developments for the method is long. For example, it can be applied to the other greenhouse gases such as CO2.

To improve the observational coverage, satellite data will be used in cooperation with module B. The initial guess for the emissions will be refined together with module Q&S. The development of the ITMS-Demonstrator will be done in module M, in comparison with different approaches to top-down inversions. As part of ITMS, the experts will be in close cooperation to compare the top-down and bottom-up approaches.Meanwhile, the ITMS-Demonstrator is an important foundation for an operational, observation-based greenhouse gas monitoring system for Germany. This will ultimately help guide climate change mitigation efforts in Germany.

Article by:

Niklas Becker, Valentin Bruch, Beatrice Ellerhoff and Maya Harms